‘Misbehaving’ RCTs: The Confounding Problem of Human Agency
Abstract
This paper argues that the theoretical model of causal inference underpinning RCTs is frequently undermined by the failure of different actors involved in their implementation to behave in ways required by the model. This is not a problem unique to RCTs, but it poses a greater challenge to them because it undercuts their claims to methodological superiority based on the ‘clean identification’ of causal effects.